COVID‐19 during the index hospital admission confers a ‘double‐hit’ effect on hip fracture patients and is associated with a two‐fold increase in 1‐year mortality risk

Abstract Purpose The aims were to: (1) determine 1‐year mortality rates for hip fracture patients during the first UK COVID‐19 wave, and (2) assess mortality risk associated with COVID‐19. Methods A nationwide multicentre cohort study was conducted of all patients presenting to 17 hospitals in March‐April 2020. Follow‐up data were collected one year after initial hip fracture (‘index’) admission, including: COVID‐19 status, readmissions, mortality, and cause of death. Results Data were available for 788/833 (94.6%) patients. One‐year mortality was 242/788 (30.7%), and the prevalence of COVID‐19 within 365 days of admission was 142/788 (18.0%). One‐year mortality was higher for patients with COVID‐19 (46.5% vs. 27.2%; p < 0.001), and highest for those COVID‐positive during index admission versus after discharge (54.7% vs. 39.7%; p = 0.025). Anytime COVID‐19 was independently associated with 50% increased mortality risk within a year of injury (HR 1.50, p = 0.006); adjusted mortality risk doubled (HR 2.03, p < 0.001) for patients COVID‐positive during index admission. No independent association was observed between mortality risk and COVID‐19 diagnosed following discharge (HR 1.16, p = 0.462). Most deaths (56/66; 84.8%) in COVID‐positive patients occurred within 30 days of COVID‐19 diagnosis (median 11.0 days). Most cases diagnosed following discharge from the admission hospital occurred in downstream hospitals. Conclusion Almost half the patients that had COVID‐19 within 365 days of fracture had died within one year of injury versus 27.2% of COVID‐negative patients. Only COVID‐19 diagnosed during the index admission was associated independently with an increased likelihood of death, indicating that infection during this time may represent a ‘double‐hit’ insult, and most COVID‐related deaths occurred within 30 days of diagnosis.


| INTRODUCTION
Patients sustaining a hip fracture are especially vulnerable to contracting and dying from Coronavirus Disease 2019  and account for a large number of inpatient deaths from COVID-19. (Clement, Ng, et al., 2020;Hall et al., 2020) Nevertheless little is known about the prevalence of infection once patients are discharged from the acute hospital stay, and the longterm effects of COVID-19 on hip fracture patients remains unclear.

Findings of previous studies by the International Multicentre
Project Auditing COVID-19 in Trauma and Orthopaedics (IMPACT) Group found that, after adjusting for confounding factors, the risk of 30-day mortality for hip fracture patients with COVID-19 was three times greater than COVID-negative patients (COVIDSurg, 2021;Hall et al., 2020;Hall & Public Health Scotland, 2021). The multinational IMPACT-Global Hip Fracture Audit demonstrated that, in the context of COVID-19 infection, increased age, male sex, frailty, and pre-existing renal and pulmonary disease were associated with an increased 30-day mortality risk (Hall, Clement, IMPACT-Global, et al., 2022;Hall, Clement, MacLullich, Simpson, Johansen, et al., 2022;. However, it is unclear whether the increased risk of death from COVID-19 is diminished once patients are past the peri-injury period, whether timing of infection relative to the hip fracture is important, and if the all-cause mortality for patients affected (at any time) by COVID-19 is significantly different to that of COVID-negative patients.
Nosocomial transmission of SARS-CoV-2 may have accounted for around half of all cases of COVID-19 around the time of hip fracture (Hall, Clement, MacLullich, Ojeda-Thies, et al., 2021;. Despite this, there is little evidence concerning the prevalence of COVID-19 in patients following discharge from hospital, or the patterns of transmission in this patient group where there is often a need for extended periods of care in inpatient hospital or rehabilitation facilities ('downstream' hospitals) or residential care settings.
The primary aim of this nationwide multicentre study was to determine the 1-year mortality rate for patients with hip fracture during the first UK COVID-19 wave. Secondary aims were to assess mortality risk associated with: (1) COVID-19 diagnosed during the initial hip fracture ('index') acute hospital admission, and (2) COVID-19 diagnosed following discharge from the acute hospital.

| METHODS
The International Multicentre Project Auditing COVID-19 in Trauma and Orthopaedics (IMPACT) Group is a clinical audit network established to facilitate investigation into the effects of the COVID-19 on hip fracture patients. (Hall, 2020 (2); Hall & Public Health Scotland, 2021) A multicentre observational cohort study was conducted of all patients presenting with a hip fracture to 17 Scottish hospitals between 1 st March and 14 th April 2020 and data were collected pertaining to patient, injury, and management factors, COVID-19 status, and outcomes. The findings, which related to risk of acquiring infection, routes of transmission, and 30-day mortality, were presented in the IMPACT-Scot Report 2 (Hall, Clement, MacLullich, Ojeda-Thies, et al., 2021;. The current study collected follow-up data up to 365 days after the date of index admission (defined as the time between initial presentation and discharge from the acute unit that provided definitive hip fracture care) and included: COVID-19 status, date of positive COVID-19 status, discharge destination, readmission to acute hospital, reason for readmission, survival status, primary cause of death, and whether COVID-19 was a contributing factor to death (defined as death within 28 days of a positive COVID-19 diagnosis, or COVID-19 listed on the death certificate) (Office for National Statistics, 2021). Data were collected by clinicians or specialist hip fracture audit coordinators local to each unit in accordance with UK Caldicott principles, and anonymised data were submitted to the central IMPACT Project Lead Team (Caldicott, 1999).

| Inclusion and exclusion criteria
The current study included all patients in the IMPACT-Scot Report 2, which applied the Scottish Hip Fracture Audit of all patients aged 50 years and over and admitted with an acute hip fracture to any of the participating hospitals over the study period (1 st March to 15 th April 2020). (Hall, Clement, MacLullich, Ojeda-Thies, et al., 2021;(SHFA), n.d.) Intracapsular and extracapsular fractures of the proximal femur up to and including the subtrochanteric region (defined as 5 cm distal to the lesser trochanter) were included. Periprosthetic fractures and isolated fractures of the public ramus, acetabulum and greater trochanter were excluded.

| Data collection
Follow-up data were collected from electronic patient records (EPR) by clinicians and specialist audit coordinators in each unit using a bespoke digital IMPACT Revisited data collection tool which utilised data-validated fields to ensure accuracy and consistency of coding, thus increasing intra-and inter-observer reliability. Data collectors had continuous access to the audit designer (AH) for queries. In addition to the verification of the original IMPACT-Scot Report 2 706 -HALL ET AL. dataset, which included demographics, injury factors, comorbidities and laboratory blood results, data were collected for all patients including: length of index acute admission; discharge destination; COVID-19 status (and date of positive any diagnosis); readmission to acute hospital (and clinical reason); mortality status, and cause of death (including primary cause, whether COVID-19 was a contributing factor, and whether there was a positive COVID-19 diagnosis within 28 days of death) (Office for National Statistics, 2021). All variables were examined up to 365 days following the index date of admission.
Continuous variables were assessed for significant differences between groups using an independent paired t-test, and categorical variables were assessed using a Chi-square test. Kaplan-Meier analysis was used to assess 365-day survival and Log rank was A Schoenfeld test determined that the proportional hazard assumptions were upheld, and Cox proportional hazard regression analysis was used to assess the independent association of COVID-19 status and 365-day mortality. A p-value of <0.05 was defined as statistically significant.

| RESULTS
One-year follow-up data were collected from 16 study centres, accounting for 788/833 (94.6%) of patients from the original study, and one study centre did not participate. There were 528 (67.0%) female and 260 (33.0%) male patients with a mean age of 80.5 years (standardised mean difference (SMD) 10.3 years). Table 1

| Prevalence of COVID-19 and mortality at one year: Univariate analyses
There were 142/788 (18.0%) patients that were diagnosed with COVID-19 within 365 days of the date of index hip fracture admission (Table 2). There were 64/142 (45.1%) patients that were diagnosed with COVID-19 during the index admission, and 78/142 (54.9%) that received a positive COVID-19 diagnosis after discharge from the acute stay. Of the 78 COVID-19 diagnoses made after the acute stay 34 (43.5%) were made within 30 days of discharge, and 25/34 (73.5%) were made in patients that had been discharged to downstream hospital settings. COVID-19 was documented as a contributing factor in 61/242 (25.2%) of deaths and was the primary cause (as stated on the Medical Certificate of Cause of Death) in 40/242 (16.5%) patients, making it the most frequent single cause of death reported. Figure 1 demonstrates the timing of positive COVID-19 cases and the duration of survival from the date of COVID-19 diagnosis.
The 1-year mortality rate for patients that were diagnosed with COVID-19 at any time was significantly higher than for those that did not have COVID-19 (46.5% vs. 27.2%, p < 0.001, log rank test ( Figure 2)). Further, the rate of mortality at one year for patients that were diagnosed with COVID-19 during the index hospital stay was significantly lower than for patients diagnosed with COVID-19 after discharge (54.7% vs. 39.7%, p = 0.025, log rank test), and mortality rates for both groups were significantly lower than for patients that did not have COVID-19 at any time (p < 0.001 and p = 0.025 respectively, log rank test ( Figure 3). The 1-year mortality rate for the 25 patients who were discharged to a downstream hospital setting and diagnosed with COVID-19 within 30 days of discharge of the acute admission hospital stay was 44.0%.

| Effect of COVID-19 on mortality at one year: Multivariate analyses
Logistic regression analysis was used to determine factors positively associated with a positive COVID-19 diagnosis. Factors that demonstrated an association (p < 0.1) with an increased likelihood of COVID-19 on univariable analysis were included in a multivariable model. Those with a significant (p < 0.05) association with a COVID-19 diagnosis on multivariable analysis included: age, male sex and higher ASA grade (Table 3). These factors were included in Cox proportional hazard regression analysis to determine the independent effect of COVID-19 status on on-year survival ( -707 T A B L E 1 Patient demographics, nottingham hip fracture score, residence, place of injury, comorbidity, anaesthesiologists (ASA) grade, COVID-19 status, and time of COVID-19 diagnosis, according to 365-day mortality

| Readmissions and associated morbidity
There were 221/788 (28.0%) patients that were readmitted to an acute hospital within 365 days of presentation with hip fracture; 56/ 788 (7.1%) patients were admitted for falls (making falls the most common reason for readmission) and in 32 cases a new fracture had been sustained (Table 2). Non-COVID respiratory system conditions were also common reasons for readmission (31/788; 3.9%), and in 9/ 788 (1.1%) cases COVID-19 was given as the reason for readmission.

| The effect of timing of a post-discharge COVID-19 diagnosis on survival
A sub-group analysis of 1-year mortality following fracture was conducted for the 78 patients that were COVID-positive after discharge from the acute stay. These patients were dichotomised into those diagnosed with COVID-19 before or after 1 st January 2021.
This date was selected because it was assumed that the majority of

| Timing of deaths among patients with COVID-19
The median survival time from initial presentation for COVID-

| Missing data
Missingness analysis revealed that American Society of Anaesthesiologists (ASA) grade was missing for 81/788 (10.2%) of patients, and of these 61 (75.3%) related to patients from two of the study centres.
Further investigation determined that this is accounted for by data collection omission relating to the unavailability of records, rather than to the respective subjects, therefore these data were handled as missing completely at random (MCAR). et al., 2020). Acute hip fracture services require the delivery of highly specialised multidisciplinary care to meet the complex needs of patients, and adherence to evidence-based national standards has been shown to be associated with better outcomes, including lower mortality rates, shorter length of hospital stay, and a higher chance of discharge back to the pre-fracture level of care (Farrow et al., 2018Metcalfe et al., 2019).

| DISCUSSION
Almost half of the cases of COVID-19 diagnosed within a year of a hip fracture were identified within the first 30 days of discharge from the acute stay, and the majority of these cases were in patients who were discharged to a downstream care facility (i.e. an inpatient or residential unit other than an acute hospital). This is a concern given the high incidence of nosocomial infection demonstrated by our previous study, which showed that around half of all COVID-19 cases diagnosed early in the patient journey (within 30 days of initial presentation) were presumed to be hospital-acquired . recent concerns about incomplete vaccination and vaccination booster programmes, as well as the emergency of SARS-CoV-2 strains that demonstrate resistance to both vaccine-mediated and previous infection-mediated immunity (Wang et al., 2021). Similar such studies are unlikely to be feasible now. Future work should utilise existing clinical audit mechanisms to monitor long-term effects on a population-wide level, and the compilation of data from multiple national hip fracture audit programmes (adhering to a minimum common clinical dataset) would ensure generalisability to a range of regions and healthcare contexts Johansen et al., 2022).

| CONCLUSION
Almost half of the hip fracture patients affected by COVID-19 had died within a year of injury and COVID-19 was a contributing factor in a quarter all observed deaths. A positive COVID-19 status at any point within a year of hip fracture was independently associated with a 50% increased risk of death in the same period, and the majority of deaths in COVID-positive patients occurred within a month of diagnosis of the disease. When adjusting for confounding factors, only a COVID-19 diagnosis made during the acute admission was associated with an increased mortality risk, and this group had a two-fold increased risk of death within a year of fracture compared to COVID-negative patients. This may reflect a 'double hit' insult in which patients are more vulnerable to contracting and dying from COVID-19 during a period of acute illness and emergency surgery.